Abstract: A touch light-emitting module having hallowed portion for incapsulation resin and a manufacturing method thereof are disclosed. The touch light-emitting module includes a printed circuit board and a touch-control conductor. The printed circuit board has a top surface on which a touch-control IC and a luminous unit that is electrically connected are disposed. The touch-control conductor includes a hollowed portion. The touch-control conductor is coated, on the bottom thereof, with a conductive material to combine with the top surface of the printed circuit board, so that the touch-control IC and the luminous unit are located in the hollowed portion. An encapsulation resin is then injected into a space between the printed circuit board and the hollowed portion to complete encapsulation. As such, the module offers a simplified structure to achieve an effect of minimization, and simplifies the manufacturing process and reduces the working time to thereby enhance the yield.

Abstract: A keyboard includes a base plate, a first key module, a second key module, a first light source, a second light source and an insulation layer. The first key module includes a first keycap disposed on the base plate, wherein the first keycap includes a first pattern aligned in a first direction. The second key module includes a second keycap disposed on the base plate, wherein the second keycap includes a second pattern different from the first pattern. The first light source is disposed under the first keycap at a position corresponding to the first pattern. The second light source is disposed under the second keycap at a position corresponding to the second patter. The insulation layer is disposed between the base plate and the first light source, wherein the insulation layer has a third pattern covering the first light source.

Abstract: Implementations of the disclosure provide systems and methods for motion compensation prediction. The method may include performing, by a video processor, a geometry partition on a video block of a video frame from a video to obtain a first partition and a second partition. The method may further include applying, by the video processor, a first motion prediction mode to the first partition and a second motion prediction mode to the second partition. At least one of the first motion prediction mode or the second motion prediction mode is an affine motion prediction mode.

Abstract: A method of manufacturing a semiconductor structure, comprising: disposing a dielectric layer over a semiconductive wafer defined with a plurality of active regions and a scribe line region surrounding each of the plurality of active regions; forming a plurality of interconnect structures within the dielectric layer, wherein the formation of the plurality of interconnect structures includes forming a plurality of first testing pads within the scribe line region and at least partially exposed through the dielectric layer; and sawing the semiconductive wafer along the scribe line region to form a first interposer and a second interposer, wherein each of the plurality of first testing pads is at least partially removed by the sawing of the semiconductive wafer.

Abstract: A computing device performs a method of decoding video data by acquiring a video bitstream including data associated with multiple encoded pictures, each picture including multiple rows of coding tree units (CTUs) and each CTU including one or more coding units (CUs). A data buffer storing a plurality of history-based motion vector predictors is used for encoding the rows of CTUs and the decoding process resets the buffer before decoding a first CU of a current row of CTUs. For a current CU of the row of CTUs, a motion vector candidate list is constructed from exploiting spatial and temporal correlation of motion vectors of neighbouring code units as well as the history-based motion vector predictors in the buffer. Finally, one motion vector predictor is selected, from the motion vector candidate list, for decoding the current CU and the buffer is updated based on the selected one.

Abstract: Implementations of the disclosure provide video processing apparatuses and methods. The method receives, by a video processor, a video block of a video for in-loop filtering. The method then performs a wavelet-domain CNN filtering on video data of at least a part of the video block, by performing, by the video processor, a wavelet transform on the video data to obtain data in a wavelet domain comprising a plurality of wavelet subbands; filtering, by the video processor, the data in the wavelet domain by applying respective CNN models on the plurality of wavelet subbands, where the CNN models for the plurality of wavelet subbands are trained in the wavelet domain; and performing, by the video processor, an inverse wavelet transform on the filtered data to obtain reconstructed video data.

Abstract: A method for removing branches from trained deep learning models is provided. The method includes steps (i)-(v). In step (i), a trained model is obtained. The trained model has a branch structure involving one or more original convolutional layers and a shortcut connection. In step (ii), the shortcut connection is removed from the branch structure. In step (iii), a reparameterization model is built by linearly expanding each of the original convolutional layers into a reparameterization block in the reparameterization model. In step (iv), parameters of the reparameterization blocks are optimized by training the reparameterization model. In step (v), each of the optimized reparameterization blocks is transformed into a reparameterized convolutional layer to form a branchless structure that replaces the branch structure in the trained model.

Abstract: A method of decoding a video signal, apparatus, and a non-transitory computer-readable storage medium are provided. The method includes obtaining a video block from the video signal, obtaining spatial neighboring blocks based on the video block, obtaining up to one left non-scaled motion vector predictor (MVP) based on the multiple left spatial neighboring blocks, obtaining up to one above non-scaled MVP based on the multiple above spatial neighboring blocks, deriving an MVP candidate list based on the coding block, the multiple left spatial neighboring blocks, and the multiple above spatial neighboring blocks by performing no motion vector scaling in all scenarios for deriving motion vectors in deriving process for spatial MVP candidates, receiving an index indicating an MVP in the MVP candidate list, and obtaining a prediction signal of the coding block based on the MVP indicated by the index.

Abstract: Methods, apparatuses, and non-transitory computer-readable storage mediums are provided for video coding with multi-direction intra prediction (MDIP). In one method, a decoder determines a blending parameter for the MDIP based a size of a coding unit. In another method, the decoder may determine a size of each coding block, and in response to determining that the size of the coding block is smaller than or equal to a first preset threshold, the decoder may determine that the MDIP mode is disabled for the coding block.

Abstract: In aspects of the disclosure, a method, a system, and a computer-readable medium are provided. The method of building a kernel reparameterization for replacing a convolution-wise operation kernel in training of a neural network comprises selecting one or more blocks from tensor blocks and operations; and connecting the selected one or more blocks with the selected operations to build the kernel reparameterization. The kernel reparameterization has a dimension same as that of the convolution-wise operation kernel.

Abstract: A passivation equipment and a passivation method for a semiconductor device are provided in the present invention. The passivation equipment for the semiconductor device includes a chamber housing and a splitter disposed in the chamber housing. The splitter divides the chamber housing to a first chamber and a second chamber. The passivation equipment further includes a first intake tube connected to the first chamber, a plasma producing unit disposed in the first chamber and a pressure detecting unit connected to the first chamber. By using the passivation equipment of the present invention, high-pressure plasma is used to increase a passivation efficiency of the semiconductor device and decrease a temperature of a passivation reaction.

Abstract: A method of subblock-based temporal motion vector prediction is performed at a computing device. The computing device acquires a video bitstream including data associated with multiple encoded pictures. While decoding a current picture in the video bitstream, the computing device selects, according to syntax elements signaled in the video bitstream, one reference picture as a collocated picture of the current picture, and determines a temporal vector between the collocated picture and the current picture from motion information of spatially neighboring blocks of a current code unit (CU) according to a fixed order. Next, the computing device splits the current CU into multiple sub-CUs, obtains a temporal motion vector predictor for each sub-CU from the temporal vector and motion information of a block in the collocated picture that corresponds to a respective subblock of the current picture and decodes the current CU according to the temporal motion vector predictors.

Abstract: A method for enhancing kernel reparameterization of a non-linear machine learning model includes providing a predefined machine learning model, expanding a kernel of the predefined machine learning model with a non-linear network for convolution operation of the predefined machine learning model to generate the non-linear machine learning model, training the non-linear machine learning model, reparameterizing the non-linear network back to a kernel for convolution operation of the non-linear machine learning model to generate a reparameterized machine learning model, and deploying the reparameterized machine learning model to an edge device.

Abstract: A reparameterization method for initializing a machine learning model includes initializing a prefix layer of a first low dimensional layer in the machine learning model and a postfix layer of the first low dimensional layer, inverting the prefix layer to generate an inverse prefix layer of the first low dimensional layer, inverting the postfix layer to generate an inverse postfix layer of the first low dimensional layer, combining the inverse prefix layer, the first low dimensional layer and the inverse postfix layer to form a high dimensional layer, generating parallel operation layers from the high dimensional layer, and assigning initial weights to the parallel operation layers.

Abstract: A method of decoding video data, comprising: receiving, from bitstream, video data corresponding to a non-skip mode coded block; determining a parameter set associated with the non-skip mode coded block; if the parameter set satisfies a first condition, retrieving a first syntax element from the bitstream; if the parameter set satisfies a second condition, retrieving a second syntax element from the bitstream; if the parameter set satisfies a third condition, retrieving a third syntax element from the bitstream; decoding the video data corresponding to the non-skip mode coded block using palette mode when the first syntax element and the third syntax element have the first value; decoding the video data using intra prediction mode when the first syntax element has the first value and the third syntax element has a second value; and decoding the video data using inter prediction mode when the first syntax element has the second value.

Abstract: A system controller includes a light source, an optical guiding assembly, a sensor and a feedback controller, and is applied to perform a solder operation with a solder module. The light source emits waveband light guided to a to-be-soldered area for heating. The optical guiding assembly is disposed between the light source and the solder module, and the waveband light is guided to the solder module by the optical guiding assembly. The sensor is disposed on another side of the optical guiding assembly for receiving a sensing light beam and then generating a sensing signal. The sensing light beam is guided to the sensor by the optical guiding assembly. The feedback controller is connected with the sensor and the light source for receiving the sensing signal and then controlling the light source. The system controller is a static part, and the solder module is a dynamic part.

Abstract: A method for video encoding is provided. The method includes: generating an inter prediction of a current coding block based on at least one motion vector from a current picture to at least one reference picture respectively; generating an intra prediction of the current coding block based on an intra prediction mode; generating a final prediction of the current coding block by weighted averaging the inter prediction and the intra prediction; determining the current coding block is treated as inter mode upon constructing a most probable mode (MPM) list of a neighboring coding block, or determining the current coding block is treated as intra mode upon constructing an MPM list of a neighboring coding block; and putting entropy-encoded information indicating that the current coding block is treated as inter mode or intra mode in a bitstream to be transmitted to a decoder.

Abstract: Various embodiments of the present disclosure are directed towards an integrated chip including a dielectric structure sandwiched between a first electrode and a bottom electrode. A passivation layer overlies the second electrode and the dielectric structure. The passivation layer comprises a horizontal surface vertically below a top surface of the passivation layer. The horizontal surface is disposed above a top surface of the dielectric structure.

Abstract: A method of decoding a syntax element for a current coding unit of video data is performed by an electronic apparatus. The electronic apparatus identifies, for the current coding unit, an above coding unit and a coding tree unit including the current coding unit. After determining that the above coding unit is within the coding tree unit, the electronic apparatus decodes, from a video bitstream, a corresponding syntax element for the current coding unit based, at least in part, on a syntax element associated with the above coding unit retrieved from a line buffer associated with the coding tree unit; otherwise, the electronic apparatus decodes, from the video bitstream, the corresponding syntax element for the current coding unit based, at least in part, on a default value assigned to the syntax element associated with the above coding unit.

Abstract: Methods and devices are provided for decoding a video block in GPM. The method includes: partitioning the video block into first and second geometric partitions; receiving a first GPM with a motion vector refinement (GPM-MVR) enable flag for the first geometric partition and receiving a second GPM-MVR enable flag for the second geometric partition; receiving a joint template matching (TM) enable flag for the first and second geometric partition that jointly indicates whether a uni-directional motion of the first partition is refined by TM and whether a uni-directional motion of the second partition is refined by the TM; receiving a first merge GPM index for the first geometric partition and a second merge GPM index for the second geometric partition; constructing a uni-directional MV candidate list of the GPM; and generating a uni-directional MV for the first geometric partition and a uni-directional MV for the second geometric partition.